A debut crime novel that attracted a reported $2 million (£1.5m) offer has been withdrawn from sale after the author’s agents said they could no longer authenticate how the manuscript was produced, marking the highest-profile publishing deal yet to collapse over suspected AI involvement.
The manuscript, Call Me, I’ll Hide the Body by Jerry Falade, was pulled by agent Marc Gerald of Europa Content after a 14-way auction and multiple six-figure bids on both sides of the Atlantic, The Guardian reported.
What happened? Falade’s manuscript was reportedly “stunningly good” — Gerald told The Bookseller that “everyone fell in love with it” when he first read it in June. The book went through a competitive auction process, attracting offers from Minotaur (owned by Macmillan US) and several UK publishers. But halfway through submissions, an editor raised concerns about potential AI use. After a meeting on July 29 where “aspects of [Falade’s] story changed,” the agency decided to cut ties.
🔍 THE BOTTOM LINE
The publishing industry has no reliable way to detect AI-written manuscripts — and the absence of that tool is now destroying deals worth millions. Authors, agents, and publishers are operating on suspicion and gut feeling in a market where AI-generated text is increasingly indistinguishable from human writing. The fallout is falling unevenly, and the pattern deserves attention.
The Publishing Industry’s AI Detection Problem
Gerald said Europa opted against using AI-detection software to check the manuscript, saying Falade was “clear and credible” when he denied using AI. That decision — to trust the author over the technology — highlights a fundamental problem: AI-detection tools are widely known to produce false positives, and the industry has not settled on a standard for vetting manuscripts.
An email from Falade’s agents to publishers, seen by the Guardian, said: “During the submission process we vetted the manuscript with Jerry for AI and felt trusting in his reassurance that it was written without utilising AI as a resource in the writing or editorial process. We can no longer substantiate that claim.”
The agents didn’t say they found evidence of AI use. They said they could no longer authenticate the process — a standard they set and then couldn’t meet.
The Racial Dimension
Falade pushed back forcefully. In a statement to the Guardian, he rejected the claims and said the response reflected racial bias in publishing:
“Three Black authors received major publishing deals this year, and all three later had deals cancelled or disrupted following suspicions about AI. There seems to be a troubling assumption that when a Black writer produces work that attracts major attention, the default is to question whether it’s really theirs.”
Falade pointed to two other Black authors — Mia Ballard and HM Wolfe — who faced similar AI-use allegations this year. Both denied using AI. Hachette previously pulled the US release of Ballard’s horror novel amid similar suspicions.
“In my case, rumours began circulating and, within six hours, I had been dropped by my agency, deals were paused, and press releases were being issued,” Falade said.
The AI Detection Industry’s Uncomfortable Truth
AI-detection tools work by analysing statistical patterns in text — word frequency, sentence structure, perplexity scores. But these tools struggle with edited prose, non-native English writing, and stylistically distinctive voices. A study published earlier this year found leading detectors flag classic literature as AI-generated and fail on carefully prompted AI text.
Gerald’s decision not to use detection software was, paradoxically, the more responsible choice — given the false positive problem. But it left the agency with no mechanism to verify Falade’s denials, and when doubts surfaced, suspicion alone became enough to collapse a deal.
This connects to broader questions we’ve tracked about AI industry workers raising safety concerns — the tools to govern AI are lagging behind the technology itself, and the gaps are producing real consequences for real people.
NZ Angle: What This Means for Local Publishing
New Zealand’s publishing industry is smaller and more concentrated than the US or UK markets, which means the AI detection problem hits differently. A single collapsed deal can represent a significant portion of an author’s annual income. NZ publishers including Penguin Random House NZ and Auckland University Press have not publicly stated their AI vetting policies. The absence of clear standards — what counts as “AI-assisted” versus “AI-generated” — is a gap that could produce its own controversies as AI tools become standard writing aids.
The question for NZ publishers is the same one facing the global industry: when a manuscript is good, who decides whether it’s too good?
❓ FAQ
Was AI use confirmed in Falade’s manuscript? No. The agents said they could no longer authenticate how the manuscript evolved, not that they found evidence of AI use. Falade denies using AI.
Why didn’t the agents use AI-detection software? Agent Marc Gerald said Falade was “clear and credible” in his denials. AI-detection tools are known to produce false positives, which makes them unreliable for high-stakes decisions.
How common are AI-related publishing deal collapses? At least three high-profile deals have been disrupted this year over AI suspicions. The Falade case is the highest-value so far, with a reported $2m offer.
What does Falade say about the racial pattern? He noted that three Black authors received major deals this year, and all three had deals cancelled or disrupted following AI suspicions — two of whom also deny using AI.
🔍 THE BOTTOM LINE
A $2m deal collapsed not because anyone found AI in the manuscript, but because the industry has no reliable way to prove it isn’t there. That’s the real story: AI detection doesn’t work well enough to be fair, and the cost of that failure falls on the people least able to absorb it. The publishing industry needs standards, not suspicion — and it needs them before more careers are damaged by a tool gap it created by adopting AI faster than it learned to police it.
